Ethanol and acetone from Douglas-fir roots stressed by Phellinus
sulphurascens infection: Implications for detecting diseased trees and
for beetle host selection
Kelsey, R. G., Joseph, G., Westlind, D., & Thies, W. G. (2016). Ethanol and
acetone from Douglas-fir roots stressed by Phellinus sulphurascens infection:
Implications for detecting diseased trees and for beetle host selection. Forest
Ecology and Management, 360, 261-272. doi:10.1016/j.foreco.2015.10.039
10.1016/j.foreco.2015.10.039
Elsevier
Version of Record
http://cdss.library.oregonstate.edu/sa-termsofuse
Forest Ecology and Management 360 (2016) 261–272
Contents lists available at ScienceDirect
Forest Ecology and Management
journal homepage: www.elsevier.com/locate/foreco
Ethanol and acetone from Douglas-fir roots stressed by Phellinus
sulphurascens infection: Implications for detecting diseased trees and for
beetle host selection
Rick G. Kelsey a,⇑, Gladwin Joseph b,1, Doug Westlind a, Walter G. Thies a
a
b
USDA Forest Service, Pacific Northwest Research Station, Corvallis, OR 97331, United States
Oregon State University, United States
a r t i c l e
i n f o
Article history:
Received 1 July 2015
Received in revised form 20 October 2015
Accepted 22 October 2015
Available online 2 November 2015
Keywords:
Laminated root rot
Phellinus weirii
Pseudotsuga menziesii
Tree stress
Root disease detection
Beetle primary attraction
a b s t r a c t
Phellinus sulphurascens (previously the Douglas-fir form of Phellinus weirii) is an important native pathogen causing laminated root rot in forests of western North America. Visual crown symptoms, or attacks
by bark or ambrosia beetles appear only during advanced stages of the disease with extensive infection in
the lower bole. Ethanol synthesis is one of many physiological responses in tree tissues stressed by pathogens. Ethanol, acetone and other volatiles from root tissues of healthy and diseased trees were analyzed
using headspace gas chromatography. Xylem and phloem from 20 diseased trees at two western Oregon
sites contained higher concentrations of ethanol, acetone, or other headspace volatiles than 20 healthy
trees on one or more dates in September, November, or the following May. Root cross-sections from eight
diseased trees were sampled along perpendicular transects and found to contain extremely variable ethanol concentrations, with highest xylem quantities in a 0–2 cm zone outside the infection boundary and
lowest amounts inside the infection. Acetone concentrations were the opposite. Logistic regression models were built and tested to determine which volatiles could predict diseased trees. A model using xylem
ethanol concentrations as the only parameter was selected and validated with measurements from 80
trees on the edges of P. sulphurascens infection centers at two different western Oregon sites. This model
successfully predicted trees with laminated root rot (78% overall correct classification and 68% for known
diseased trees), but worked best for those with infections observed in both root cores and the root collar
(100% correct). Early detection of P. sulphurascens infected trees remains a challenge. Our ethanol analysis
method is useful for research, but provides limited benefits for identifying individual P. sulphurascens
hazard trees, or for extensive ground surveys in the forest. Whether ethanol is released to the atmosphere
in sufficient quantities to confirm infection before the late appearance of crown symptoms, or bark
beetles remains unknown. If it is, then development of sensors capable of tree side detection requiring
minimal tissue sampling would be useful in managing this disease. We also propose a mechanism for
how ethanol with host monoterpenes could play a central role in pioneering bark beetle primary host
selection of trees infected with this pathogen.
Published by Elsevier B.V.
1. Introduction
When trees are subjected to stress from various biotic or
abiotic agents one of their many physiological responses may
be ethanol synthesis if the stressed cells experience impaired
aerobic respiration as shown when tissues are deprived of oxygen
(Kelsey et al., 1998a; Joseph and Kelsey, 2004). This allows the
⇑ Corresponding author.
E-mail addresses: rkelsey@fs.fed.us (R.G. Kelsey), gladwin.joseph@apu.edu.in (G.
Joseph), dwestlind@fs.fed.us (D. Westlind), wgthies@gmail.com (W.G. Thies).
1
Current address: Azim Premji University, Bangalore, India.
http://dx.doi.org/10.1016/j.foreco.2015.10.039
0378-1127/Published by Elsevier B.V.
cells to survive for brief periods until their O2 supply is
restored, and if not restored they die. Ethanol accumulation is
dependent on the duration and rates of synthesis (Kelsey
et al., 2011), its subsequent dissipation by diffusion as shown
by adding it to tissues (Kelsey et al., 2013), movement in the
transpiration stream (Kreuzwieser et al., 1999; Cojocariu et al.,
2004), metabolism in non-stressed cells to other cellular components (MacDonald and Kimmerer, 1993), direct release to the
atmosphere (MacDonald and Kimmerer, 1993; Kreuzwieser
et al., 1999; Rottenberger et al., 2008; Ranger et al., 2013), or
conversion to acetaldehyde that is released to the atmosphere
(Cojocariu et al., 2004).
262
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
Pathogens may be the most common biotic stress agents to
induce ethanol synthesis in trees. For example, boles of
lodgepole pine, Pinus contorta Douglas ex Loudon, with active
decay fungi emit 2.4 times more ethanol to the atmosphere than
those without the fungi (Gara et al., 1993). Elevated ethanol
concentrations occur in stem tissues near the root collar of
Douglas-fir, Pseudotsuga menziesii (Mirb.) Franco., infected with
Leptographium wageneri var. pseudotsugae Harrington & Cobb,
the cause of black stain root disease (Kelsey and Joseph, 1998).
This was also observed in stem tissues of ponderosa pine, Pinus
ponderosa Lawson & C. Lawson, infected with L. wageneri var.
ponderosum (T.C. Harr. & F.W. Cobb) T.C. Harr. & F.W. Cobb, or
its combination with Heterobasidion irregular Garbelotto &
Otrosina, the cause of Heterobasidion root disease of pine
(Kelsey et al., 1998b, 2006).
Ethanol can also accumulate within the boundaries of cankers
caused by Phytophthora ramorum Werres, De Cock & Man in’t Veld,
the microbe responsible for sudden oak death, on the stems of
coast live oak, Quercus agrifolia Nee, whereas concentrations in tissues outside the cankers and in adjacent healthy trees remain low
(Kelsey et al., 2013). In the absence of any abiotic stressors, especially flooding as it induces synthesis in the roots and transport
into the bole (Kreuzwieser et al., 1999; Cojocariu et al., 2004;
Ranger et al., 2013), elevated ethanol concentrations in a tree can
be a strong indicator of infection. However, it is important to note
that low ethanol concentrations do not explicitly confirm the tree
is disease free because of the many interacting factors influencing
the rates of synthesis and subsequent dissipation. If ethanol
escapes from stressed trees into the atmosphere in sufficient
quantities and duration it can function, usually in combination
with other volatiles released from the tree, as a signal that
attracts various bark or ambrosia beetles to land and attack
(Kelsey and Joseph, 2001, 2003; Kelsey et al., 2013, 2014; Ranger
et al., 2013).
Other compounds occur with ethanol in the headspace analysis
of tree tissues and their concentration changes associated with
pathogen infections may also function as biomarkers for detecting
disease. Acetaldehyde, acetone, methanol, and propanol were all
more strongly correlated than ethanol with the severity of black
stain and Heterobasidion root diseases in ponderosa pine (Kelsey
et al., 1998b). Acetaldehyde was selected as the single best predictor of black stain disease, followed by acetone. In a related study,
284 ponderosa pine trees were randomly selected in stands where
crown symptoms were unreliable for identifying diseased trees
(Kelsey et al., 2006). In this case, sapwood acetone concentrations
were selected as the best chemical headspace markers for predicting root disease. We suspected ethanol, or one of the other headspace compounds might also serve as a useful marker for earlier
and more accurate detection of trees with laminated root rot
caused by Phellinus sulphurascens Pilát (previously the Douglas-fir
form of P. weirii).
P. sulphurascens infects various conifer hosts, with Douglas-fir,
true firs and mountain hemlock being most susceptible (Thies
and Sturrock, 1995). It spreads across root contacts between
healthy and diseased trees or stumps, with ectotrophic mycelium
transfers considered more important than endotrophic transfers
(Bloomberg and Reynolds, 1982). Because of root-to-root spread,
multiple trees often die near one another creating gaps in the forest canopy with standing dead and fallen trees. These gaps are
easily recognized signatures of this pathogen that slowly expand
outward in an uneven radial pattern at a rate of less than 50 cm
per year (Nelson and Hartman, 1975; McCauley and Cook, 1980).
However, some diseased trees also occur sporadically throughout
a stand, unassociated with gaps. The infection mechanism for these
trees is yet unknown and their detection is challenging (Thies and
Nelson, 1997).
Distinct visual symptoms often do not appear in the crowns of
infected trees (Wallis and Reynolds, 1965; Bloomberg and Wallis,
1979; Wallis and Bloomberg, 1981; Thies, 1983), and they are typically not attacked by bark beetles (Buckland et al., 1954; Lane and
Goheen, 1979; Goheen and Hansen, 1993), until the infections
reach advanced stages in the bole. Reduced annual height growth
can be an early crown symptom for P. sulphurascens (Bloomberg
and Wallis, 1979), but this change may be gradual, or variable,
and difficult to recognize when an infected tree is among healthy
ones whose height growth may also be impacted by competition,
water deficits, or other stresses. Trees with healthy appearing
crowns and no evidence of beetle colonization may be windthrown during storms exposing the broken roots weakened by
decay. These trees are hazardous when growing in campgrounds,
along road right-of-ways, or near homes. Their detection and
removal is critical. Because of its combined economic and ecological importance in Washington forests, P. sulphurascens was
recently chosen over other root pathogens as best suited for
focused research directed toward improving applied management
options (Cook et al., 2013). Their report cites methods for detecting
infected stands and conducting ground surveys, but emphasizes
the importance for further research to improve early detection,
identification accuracy, and cost-effectiveness.
Objectives for the study we report here were to: (1) determine
whether roots of Douglas-fir stressed by P. sulphurascens infections
synthesize and accumulate higher ethanol concentrations than
roots from uninfected trees as observed for other pathogens; (2)
determine if there are differences in the quantities of acetaldehyde,
acetone, methanol, or propanol detected during headspace analysis
of ethanol between roots of infected and uninfected trees; and (3)
attempt to develop a predictive model using these compounds to
help identify trees infected with P. sulphurascens before the disease
reaches an advanced stage.
2. Methods and materials
2.1. Study sites and tree selection
Four sites were utilized in this study; sites 1 and 2 were used to
gather data for model building, while sites 3 and 4 were used to
gather data for testing the model. Trees selected for model development were from two Douglas-fir stands near Philomath, Oregon.
Site 1 was at 44.475461°, 123.430978° (202 m elev.) and site 2 at
44.545358°, 123.497525° (268 m elev.). At each site 10 tentatively diseased trees were selected and tagged as encountered
when they fit the selection criteria of (1) being located on the
perimeter of a P. sulphurascens infection center; (2) being a dominant or co-dominant tree; (3) having P. sulphurascens ectotrophic
mycelium present on the bark surface of one or more partially
excavated roots; and (4) having no evidence of bark or ambrosia
beetle attacks. Visible crown symptoms for these trees were variable with some suggesting advanced infections. Rot or the characteristic stain from infection (Thies and Sturrock, 1995) was
detected in at least one root from 13 of the 20 trees considered diseased. Ten putatively healthy trees were also selected and tagged
as encountered when they fit the selection criteria of (1) being
located well beyond root contact with trees on the infection center
perimeter; (2) being a dominant or co-dominant tree; (3) having
no ectotrophic mycelium on the bark surface of one or more partially excavated roots; and (4) having no evidence of bark or
ambrosia beetle attacks. No rot or stain was observed in their sample cores on any dates. As described later, the phloem and xylem
from root cores of these trees were analyzed for headspace volatile
concentrations and the values were used to build a model for predicting whether trees at the sites below were likely infected with
this pathogen.
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
An additional 40 trees were sampled from each of two stands
near Forest Grove, Oregon for validation of the model developed
from two sites above. Site 3 was located at 45.450608°,
123.345275° (611 m elev.) and site 4 at 45.767447°,
123.353411° (375 m elev.). These trees were selected and tagged
as encountered from around the periphery of P. sulphurascens
infection centers marked for harvest, without knowing if they were
diseased or healthy. Four main roots entering the root collar near
the soil surface was the only selection criterion.
2.2. Root sampling for headspace volatile analysis and water contents
At sites 1, 2, the root cores were collected from healthy and diseased trees in September, November and the following May to
detect any seasonal changes in the volatile concentrations. Four
main lateral roots were partially exposed from the soil and tagged.
The following day (27 September) a tissue core was extracted with
an increment borer (5 mm i.d.) from the top-right side (facing the
bole) of each root at a distance of 5–30 cm below the point where
they entered the surface litter layer or soil. The phloem and first
1.0 cm of xylem were separated, sealed in glass vials
(15 45 mm o.d.) then immediately frozen with dry ice until
stored in a 36 °C freezer. Root holes were plugged with corks to
minimize further contamination and improve visibility for future
sampling. At site 1, one additional sample on diseased trees only
was collected from the lower bole above each sampled root at a
distance of 5–20 cm above the soil surface. Roots were resampled
on the top left-side at the same distance from the bole on 28 or
29 November and finally about 3–5 cm down the root from previous sample holes on 23 May to evaluate seasonal changes in ethanol concentrations.
Trees at sites 3–4 were sampled one year later, 16 and 23 May
for site 3, and 5 June for site 4. Four roots per tree were partially
excavated and one increment core (0.5 6 cm xylem depth)
extracted from the top of each at 5–30 cm distal to their point of
entry into the soil litter layer, or soil and processed as above.
Xylem depth was increased to 6 cm rather than 1 cm at sites 1,
2. This would contain the same tissue as a 1 cm core, but for larger
roots might improve the likelihood of reaching tissue near an infection with higher ethanol concentrations, as shown by the root transect experiment described below. Any staining in the xylem core
was noted as an infection. These trees were then harvested, their
tagged stumps relocated and the presence of any brown stain on
the cut surface recorded.
2.3. Headspace and water content analysis
Headspace volatiles in the samples from sites 1, 2 and root transect were processed and analyzed with the gas chromatography
(GC) instruments and settings previously described in detail
(Kelsey and Joseph, 1998). Briefly, the samples were thawed on
ice, weighed into headspace vials, and then sealed with a septum
and crimp cap (PerkinElmer, Akron, OH, USA). Vials were held on
ice until a full set was processed then placed in a 102 °C for
30 min to deactivate enzymes in the tissue. The GC was a Hewlett
Packard 5890 with flame ionization detector and J&W DB-Wax column (30 m 0.32 mm i.d., 0.25 lm film thickness) using helium as
the carrier gas. The GC injector was set at 50 or 60 °C and the detector at 250 °C. The HS40 settings were 90, 100 and 60 °C for the
sample heating block, needle, and transfer line temperatures, and
1.0, 0.04, and 0.4 or 0.1 min for the vial pressurization, injection,
and needle withdrawal times, respectively. The column oven was
held isothermally at 50 °C for 4.5 min for sites 1, 2, and the root
transect samples. For samples from sites 3, 4 the column oven
was held at 50 °C for 1.0 min, then increased to 80 °C at 20 °C per
min. and held for 1.5 min to minimize monoterpene carryover
263
between samples. Duplicate vials, each containing 5 lL of a mixed
acetaldehyde and ethanol standard were included with each sample set for quantification by the external standard method using a
linear response curve from zero for both compounds. All samples
were analyzed twice by the multiple headspace extraction technique for calculating concentrations of acetaldehyde and ethanol
(Kolb et al., 1984). Because acetone, methanol and propanol are
generated during heating (Kelsey et al., 1998b) their concentrations were determined by static headspace calculations with values from the first analysis only and those of the ethanol
standards using a relative response factor of 1. Ethanol concentrations, except for root transect samples, were calculated as lg g1
fresh mass because of its strong association with water by hydrogen bonding. Concentrations of the other compounds were
reported in the same units for consistency. Tissue water content
was measured on these samples after analysis by removing the
caps and heating at 102 °C for 16 h, then cooling in a sealed container with desiccant for 30 min before taking a final weight.
2.4. Root transect sampling and analysis
This experiment was setup to examine concentrations of headspace volatiles in P. sulphurascens infected root tissues and compare them with quantities in the surrounding healthy tissues.
One root on each of eight diseased Douglas-fir stumps from trees
harvested the previous 30–45 days near site 1 was collected
between 7 and 16 December. Diseased trees were identified by
the characteristic brown P. sulphurascens stain visible on the stump
surface and selected as encountered. One root beneath the stained
stump surface was excavated, scribed on top for future reference,
then severed 15–232 cm from the bole, and again at 1 m or less
distal from the first cut. Each root segment (dia 6–31 cm; mean
12 cm) was placed in a plastic bag to minimize water loss and
returned immediately to a 5 °C coldroom for storage and processing. Each root was first cut perpendicular to length near the proximal end and the xylem examined for a distinct zone of rot or stain
surrounded by apparently healthy tissue. If this infection pattern
was observed a disk (about 10 mm thick) was removed, but if there
was decay with loss of tissue strength and integrity, or no stain,
then a new cut was made further down the root. When disks
matched the selection criteria, the stained-infected area was outlined with a marker for future reference. A transect (0.5 cm wide)
was marked across the cut surface from root top to bottom, and
another left to right with both passing through the disk center.
Segments 10 mm long were marked and numbered along each
transect starting at the center and progressing outward through
the phloem. Xylem segments adjacent to the phloem were of variable length.
After marking, disk surfaces were photocopied for measuring
segment distances from stain or decay. The transect segments
(about 1 1 cm) were then cut out on a band saw and held in vials
on ice. Each segment was split into smaller pieces then quickly
placed into a pre-weighed headspace vial, sealed with a septum
and held on ice until all segments were processed. The vials were
heated for 60 min in a 102 °C oven to deactivate enzymes and prevent further ethanol synthesis. Volatiles and tissue dry mass were
analyzed as described above, except the samples were heated
60 min in the HS40 autosampler because of their larger mass. Volatile concentrations are reported as lg g1 dry mass, because tissue
fresh mass was not measured to facilitate faster sample processing.
To verify the pathogen was P. sulphurascens, a second disk
immediately adjacent to the one used for headspace analysis was
cut and marked with the same transect for six of the eight roots
processed. Each 10 10 mm segment of tissue was cut from the
transect, noted for presence of stain, then split into three pieces
and each placed separately on 1.5% malt extract agar slants in
264
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
culture tubes. The tubes were incubated for 21 days at 20 °C then
examined for the presence of P. sulphurascens setal hyphae in
developing colonies. Three samples of decayed tissue from one root
cross section were also cultured.
2.5. Pre-dawn water potentials at sites 1, 2
Pre-dawn water potential was measured on all trees to determine whether the pathogen infection caused greater water stress
in diseased trees than healthy trees. The first measurements were
taken at the end of summer on 28 and 29 of September and again
on 30 November and 1 December, after the fall rain had begun. A
branch tip was detached from the crown and a secondary lateral
removed for measurements with a Scholander pressure chamber
(PMS Instruments, Albany, OR, USA). Heights of the sampled
branches were measured and used to adjust for gravitational
effects on water potentials.
2.6. Diameter and radial growth measurements
Tree diameters at breast height (DBH) were measured at all
sites with a diameter tape at 1.4 m above the forest floor. The last
five-year radial growth at sites 1, 2 was measured to the nearest
mm on one increment core per tree at BH on the uphill side of
the bole. At sites 3, 4, five-year radial growth was measured to
the nearest mm at two random positions on each stump surface
and averaged, except for one stump damaged during harvest.
2.7. Statistical analysis
Although each compound was analyzed in the phloem and
xylem of four roots per tree on each sampling date, only the maximum concentration among the four roots was assigned to each
tissue on each date for analysis. A tree was categorized as diseased
regardless of the number of infected roots, and this helped to minimize any dilution effect that might occur from averaging low concentrations in three healthy roots with a high concentration in one
diseased root. The maximum concentration for all compounds did
not necessarily occur in the same root within a sampling date, nor
did the maximum value for each compound always occur in the
same root across dates. Alternatively, the water content in each tissue was averaged across the four roots of a tree on each sample
date to help mitigate the influence of a single low value among
the four roots, since water can be influenced by various factors
other than pathogen stress.
All analyses were done with the linear mixed procedure in SAS
9.4 (SAS, 2012). Assumptions of normality and constant variance of
the residuals was checked for all models using quantile–quantile
plots of the residuals and residual vs fitted value plots, respectively. If assumptions were not met, the response was natural log
transformed and assumptions were checked again to ensure they
were met. Estimates and confidence intervals for any transformed
responses were back-transformed and inference made to this median response. The Kenward–Roger method was used for calculating
degrees of freedom (Littell et al., 2006). For each analysis of
repeated measures data we fit models with compound symmetry
(CS), unstructured (UN1-3), Toeplitz (TOEP1-3), auto-regressive
(AR1) and spatial power (SP) covariance structures and chose a
final one based on the suitability with our repeated sampling and
the lowest AICc values. Comparisons among group levels were calculated for each model using protected Fisher’s least significant
difference to guard against error from multiple comparisons. All
statistical testing was done using an alpha of 0.05.
For diseased trees at site 1, the highest ethanol concentrations
per tree from the four bole samples were compared with the highest quantity among the four roots using a paired t-test for the
phloem and xylem separately after transforming the data to natural logarithms. The relationship between bole and root ethanol
concentrations in each tissue were evaluated by Pearson correlation coefficients using all four bole and root samples per tree.
Tissue mean water contents and maximum volatile concentrations for trees at sites 1, 2 were compared between the healthy
and diseased trees by a randomized block (site) repeated measures
(month) analysis for each tissue separately with the UN(1) covariance structure used in all repeated measures analyses. Main effect
of tree disease condition was fixed and site and tree were random
effects. At sites 3, 4 the mean root water content and maximum
ethanol and acetone concentrations were analyzed separately for
phloem and xylem with tree condition (diseased or healthy) as a
fixed effect and site as a random effect.
Transect sampling of eight diseased roots was used to examine
volatile concentrations inside and outside the infection and determine how proximity to the infection boundary influenced concentrations. For each tissue, transect segments were assigned to one of
five categories based on their distance from the nearest xylem
infection boundary; (1) segment partially or entirely stained, (2)
0–2 cm, (3) 2–4 cm, (4) 4–6 cm and (5) 6+ cm from the nearest
stain border (not included in the analysis because of low numbers).
A mean compound concentration was determined from the pool of
cores in each category and these analyzed separately for the xylem
and phloem in a randomized complete block (root) repeated measures (distance) design with distance a fixed main effect and root a
random effect. The SP and TOEP (1) covariance structures were
used for repeated measures analysis of the phloem and xylem volatile concentrations, respectively.
Tree water potentials at sites 1, 2 were compared using a randomized complete block (site), repeated measures (month) analysis using UN(1) covariance structure, with tree disease condition
and sampling month as fixed main effects, and site as a random
effect. Tree diameter and five-year radial growth were each combined across sites 1–4 for the diseased and healthy trees, then
compared using a randomized complete block (site) analysis, with
tree disease condition a fixed main effect, and site a random effect.
2.8. Model building with tissue measurements at sites 1, 2
Logistic regression models for predicting the probability that a
tree is diseased were created from compounds measured in the
root xylem and phloem from the 20 putative healthy and 20 diseased trees at sites 1, 2 using the SAS 9.4 logistic procedure. For
each tissue the binary response variable was tree disease condition, with mean water content and the maximum root concentrations for acetaldehyde, acetone, ethanol, propanol, and methanol
tested as explanatory variables. The tissue mean water content
was an average for all four roots across the three sample dates.
The maximum value for each compound was obtained by calculating an average concentration for each of the four roots using the
measurements from the three sampling dates, and then using the
highest average value from among the four root values to represent
the tree. Models were built using the purposeful selection of
covariates method as described by Hosmer et al. (2013). The fit
of each model to the data was checked during this process using
a receiver operating characteristic (ROC) area under the curve
(AUC), and the Hosmer–Lemshow goodness-of-fit tests.
2.9. Model validation with measurements at sites 3, 4
The model developed from sites 1, 2 was tested by comparing
its predicted probability of infection to the actual infection
observed in 80 independent trees at sites 3, 4; 25 known infected
plus 55 presumed uninfected. Three methods were used to evaluate fit; (1) ROC AUC, (2) Hosmer–Lemeshow goodness-of-fit tests,
265
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
3. Results
Table 1
Sites 1, 2 statistical analysis results for ethanol and acetone contents in root phloem
and xylem from diseased and healthy trees.
a
b
Xylem
d.f.
F
Pb
d.f.
F
Pb
Tc
M
Tc M
1, 35.3
2, 49.1
2, 49.1
17.20
2.02
0.67
<0.001
0.143
0.515
1, 46.7
2, 46.3
2, 46.3
41.33
6.96
6.20
<0.001
0.002
0.004
Acetone
Tc
M
Tc M
1, 37.0
2, 42.1
2, 42.1
21.53
0.31
0.06
<0.001
0.733
0.945
1, 33.8
2, 38.4
2, 38.4
24.91
0.64
2.55
<0.001
0.534
0.092
Tc = tree condition (diseased, healthy), M = month.
Bold P values are statistically significant.
60
50
Phloem A
Healthy
Diseased
Effect
d.f.
t
P
SH v SD 41.6 3.57 0.010
NH v ND 41.6 3.01 0.045
MH v MD 32.8 1.94 0.393
SD v ND 40.2 1.05 0.897
40
SD v MD 55.5 2.21 0.252
ND v MD 57.4 1.11 0.874
30
SH v NH 40.2 0.68 0.934
SH v MH 55.5 0.61 0.990
b
20
NH v MH 57.4 0.07 1.000
bd
10
bc
acd
ac
acd
0
Xylem B
60
Ethanol, µg.g-1 fresh mass
Ethanol concentrations in root phloem were dependent on tree
condition only (Table 1) with 12.9 lg g1 fresh mass in diseased
tree roots that was 3.2 (95% CI 1.8, 5.6) times higher than in
healthy trees. However, monthly values are presented in Fig. 1A
for consistency with xylem results. In xylem there was a tree condition by month interaction (Table 1), nevertheless at each month
the xylem ethanol in diseased trees was greater than in healthy
trees (Fig. 1B). This interaction was caused by declining quantities
in diseased trees at each subsequent sampling date that were all
different from one another, whereas in healthy tree roots the concentrations remained the same among all months (Fig. 1B). The largest difference observed for diseased xylem over healthy xylem
occurred in September roots (22.9, 95% CI 11.3, 46.4 times) and
smallest difference in May roots (4.1, 95% CI 1.3, 13.5 times).
For diseased trees only at site 1 (not graphed) the mean phloem
ethanol in the bole (12.2, ±SE 1.1 lg g1 fresh mass) was lower
than in the roots (93.6, ±SE 46.3 lg g1 fresh mass), but not statistically different (t9 = 2.06, P = 0.070), whereas xylem ethanol concentrations in the bole (28.2, ±SE 13.9 lg g1 fresh mass) were
statistically lower (t9 = 2.35, P = 0.043) than in the roots (76.9,
±SE 22.2 lg g1 fresh mass). Nevertheless, the maximum concentration measured in bole xylem (144.2 lg g1 fresh mass) was
nearly as high as in the root xylem (167.0 lg g1 fresh mass). Also,
bole xylem concentrations had a strong, positive correlation
(r = 0.729, n = 39 one outlier removed, P < .001) with quantities in
the root xylem, whereas there was no correlation (P = 0.105)
between quantities in the bole and root phloem.
Acetone concentrations in root phloem and xylem (Fig. 2) were
dependent on tree condition, but not month or their interaction
(Table 1). The 9.1 lg g1 fresh mass of acetone in the phloem of
Phloem
Ethanol
Results for the initial statistical model analyses are reported in
tables, with graphs presented for the most relevant variable comparisons. Additional results for headspace volatiles other than
ethanol and acetone, some water content information, tree diameters, and growth are presented in Appendix A.
3.1. Sites 1, 2 root ethanol and acetone concentrations
Effecta
Volatile
Ethanol, µg.g-1 fresh mass
and (3) classification tables using a probability cut-point of 0.40
(Hosmer et al., 2013). ROC AUC assesses the models ability to discriminate between infected and uninfected trees in a range from
0.0 to 1.0, with values 60.5 indicating no discrimination, P0.7 as
acceptable, P0.8 as excellent, and P0.9 as outstanding. The Hosmer–Lemeshow test partitions the trees into 10 percentile groupings from low to high probability and compares the model’s
expected number of infected with the actual observed infected. A
Pearson chi-square test is then done on a 10 2 table of the
expected and observed frequencies, with a chi-square P-value
greater than 0.05 indicating good model fit. Classification tables
compare the observed versus predicted outcomes by selecting a
cut-point probability; in this case we chose 0.40 because it maximized the correct calls for both healthy and infected. If the estimated probability was equal to or exceeded the cut-point the
tree was predicted to be infected, otherwise it was predicted to
be uninfected.
Finally, the trees from sites 3, 4 were sorted into four categories
based on observed infections: (1) stump surface only, (2) root core
only, (3) both stump surface and root core, and (4) no staininfection observed anywhere. Then the number of correct and
incorrect model predictions were tallied for the trees in each category and converted to percentages to evaluate what infection types
the model identified best.
Effect
d.f.
t
P
SH v SD 38.4 8.97 <0.001
50
NH v ND 31.3 5.05 <0.001
MH v MD 49.5 2.39 0.021
40
SD v ND 48.0 2.62 0.012
b
SD v MD 35.7 4.76 <0.001
ND v MD 50.7 2.74 0.009
30
SH v NH 48.0 0.98 0.334
SH v MH 35.7 0.54 0.593
20
c
NH v MH 50.7 1.12 0.267
10
d
ad
a
0
Sept
a
Nov
May
Month
Fig. 1. Sites 1, 2 backtransformed mean ethanol concentrations (95% CI) in Douglasfir root phloem (A), and xylem (B). See Table 1 for model P values; with P values
listed in graphs for the most biologically relevant comparisons, S = Sept, N = Nov,
M = May; H = healthy, D = diseased. Bold P values are statistically significant.
diseased roots was 3.9 (95% CI 2.1–7.0) times higher than in roots
of healthy trees (Fig. 2A). Root xylem from diseased trees contained
10.5 lg g1 fresh mass acetone that was 3.3 (95% CI 2.0, 5.4) times
greater than in healthy trees (Fig. 2A).
3.2. Sites 1, 2 root transect sampling for ethanol, acetone, and
pathogen identification
P. sulphurascens was confirmed in culture from five of the six
diseased roots sampled, but only from the visually stained tissue.
The fungus was cultured also from three samples of xylem with
advanced decay removed from one root.
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
Acetone, µg.g-1 fresh mass
266
15
Phloem A
10
b
Volatile
Ethanol
Acetone
a
5
Effect
Distance
Distance
Phloem
Xylem
d.f.
F
P
d.f.
F
Pa
2, 1.41
2, 8.88
8.60
1.98
0.162
0.195
3, 10.2
3, 11.7
10.2
4.18
0.042
0.005
Bold P values are statistically significant.
a
0
Xylem B
15
Acetone, µg.g-1 fresh mass
Table 2
Statistical analysis results for ethanol and acetone in transect segments of phloem
and xylem at different distances from the stained-infection boundary in diseased
roots.
b
10
5
a
c
0
Healthy
Diseased
Tree condition
Fig. 2. Sites 1, 2 backtransformed mean acetone concentrations (95% CI) in
Douglas-fir root phloem (A) and xylem (B) from diseased and healthy trees. Bars
with the same letters are not statistically different, see Table 1 for P values.
Xylem ethanol concentrations depended on the tissue
distance from the pathogen (Table 2, Fig. 3) with highest
amounts in the 0–2 cm zone immediately adjacent to the
stained-infected tissue that contained the lowest quantities.
Xylem 2–4 cm and 4–6 cm away from the infection boundary
had intermediate quantities of ethanol that were not statistically different from one another or the 0–2 cm and infected
tissues. Phloem at 0–2 cm from the xylem infection contained
over three times the amount of ethanol as phloem at 2–4 cm,
but there were no statistical differences among any of the three
phloem distances (Table 2, Fig. 3).
Xylem acetone concentrations were also dependent on the
tissues distance from the infection (Table 2, Fig. 3), but the
opposite of ethanol with quantities in infected tissue statistically
greater than those in adjacent healthy xylem 0–2, or 2–4 cm away.
Acetone in xylem at 4–6 cm was not statistically different from the
amounts in any of the other distance categories. Phloem acetone
concentrations also did not differ statistically among samples
collected at three increasing distances from the stained xylem
(Table 2, Fig. 3).
3.3. Sites 1, 2 pre-dawn water potentials, root tissue water contents,
and precipitation
Pre-dawn water potentials of diseased trees were similar to
healthy trees indicating that the infected trees were not experiencing additional water stress over healthy trees (Fig. 4A, Table 3). All
trees regardless of their condition were experiencing some water
stress in September near the end of summer, but it had been eliminated by November with the fall rain (Fig. 4B). There was no tree
condition by month interaction (Table 3).
Mean phloem water content in roots was dependent on the
month sampled only and not the tree condition or their interaction
(Appendix A, Table A1, Fig. A1). Phloem water contents were lowest in September, intermediate in November and highest in May
with quantities at all dates statistically different from one another
(Appendix A, Fig. A1C). Alternatively, the xylem water contents
were not different between diseased and healthy trees, among
months, or their interaction (Appendix A, Table A1, Fig. A1).
Precipitation measured between the study sites (Corvallis
Water Bureau station, 44.508°, 123.458°, elev. 180.4 m) was
3.0 cm from 1 July to 27 September, 43.6 cm from 27 September
to 27 November, and 125.3 cm from 27 November to 23 May.
3.4. Sites 3, 4 root ethanol, acetone, and water contents
The root phloem ethanol contents at sites 3, 4 were similar for
diseased and healthy trees (Table 4, Fig. 5A). However, the xylem
from diseased trees contained 8.7 lg g1 fresh mass of ethanol,
or 11.8 (95% CI 4.3, 32.4) times more than in the xylem from
healthy trees (Table 4, Fig. 5B). Root acetone concentrations in
the phloem and xylem of diseased trees were statistically greater
than in healthy trees (Table 4, Fig. 6). The 6.3 lg g1 fresh mass
in phloem was 2.1 (95% CI 1.4, 3.2) times greater than in healthy
trees, while the 2.3 lg g1 fresh mass in xylem of diseased trees
was 3.1 (95% CI 2.2, 4.4) times greater than in healthy trees. Mean
phloem water contents in trees at sites 3, 4 were similar between
diseased trees with 1.25 (95% CI 0.91, 1.57) g g1 dry mass and
healthy trees with 1.26 (95% CI 0.72, 1.81) g g1 dry mass (Table 4).
However, the 0.81 g g1 dry mass of water in xylem from diseased
trees was statistically lower (Table 4) than the 0.98 g g1 dry mass
in healthy tree roots.
3.5. Model building with root measurements at sites 1, 2
The xylem volatile concentrations measured in forty trees from
sites 1, 2, twenty each presumed healthy and diseased, were used
to develop separate logistic regression models by tissue type to
predict tree disease condition. Xylem explanatory variables
included in the initial univariate models were mean water content
and maximum values for acetone, ethanol, and methanol. Purposeful selection of xylem covariates resulted in a final model containing ethanol and acetone concentrations as significant (P = 0.076
and P = 0.089 respectively) predictors of infection by P. sulphurascens. This model was a good fit to the data based on a ROC AUC
value of 0.98 (95% CI 0.94, 1.0) and a Hosmer–Lemeshow
X26 = 3.45, P = 0.751. In an effort to produce the simplest model possible we also examined the univariate models for both xylem ethanol
and acetone. They each had good fit to the data as well, both with
ROC AUC values of 0.96 (95% CI 0.91, 1.0) and Hosmer–Lemeshow
X27 = 4.79, P = 0.685 and X27 = 7.25, P = 0.404, respectively. These models were then further validated based on their ability to predict diseased and healthy trees with the ethanol and acetone concentrations
measured in roots of trees at sites 3, 4. None of the phloem covariates from sites 1, 2 met the thresholds for inclusion in final models
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
267
Fig. 3. Backtransformed mean ethanol (bold) and mean acetone (italics) concentrations (lg g1 dry mass) in xylem and phloem at various distances from P. sulphurascens
stain-infected xylem in Douglas-fir root cross-sections (n = 8) sampled by perpendicular transects through the disks. Within each tissue-compound category those values
followed by the same letters are not statistically different. Bold P values are statistically significant.
during the purposeful selection stage. To be sure nothing was overlooked we again tested models using sites 1, 2 acetone and ethanol
concentrations alone and in combination.
3.6. Model validation and selection with root measurements at
sites 3, 4
The xylem models above were independently tested for predicting infection using the 80 trees from sites 3, 4. The univariate
model with ethanol concentrations performed the best overall
with an ROC AUC of 0.83 (95% CI 0.74, 0.92) indicating excellent
fit and a Hosmer–Lemeshow X28 = 10.51, P = 0.231:
The model with both acetone and ethanol as covariates had an
ROC AUC of 0.87 (95% CI 0.78, 0.95) indicating excellent fit and a
Hosmer–Lemeshow X28 = 13.06, P = 0.110, but none of the probability cut-points exceeded 53% correct prediction for known diseased
trees. The acetone only model had an ROC AUC of 0.84 (95% CI
0.74, 0.95) indicating excellent fit and a Hosmer–Lemeshow
X28 = 12.52, P = 0.129, but here again no cut-point exceeded a 40%
correct prediction for known diseased trees. Phloem models were
also tested with the sites 3, 4 validation data but none of them fit
particularly well, with ROC values all <0.72 and overall correct classification rates of <65.0%.
4. Discussion
LOGIT PðiÞ ¼ 2:32 þ 0:3597 ðroot xylem maximum ethanol lg g1 fresh massÞ
PðiÞ ¼ EXPfLOGITðpÞg=½1 þ EXPfLOGITðpÞg
4.1. Ethanol concentrations
Using a 0.40 cut-point for the probability of infection resulted in
a 77.5% overall correct classification across all trees and a 68.0%
correct prediction for known diseased trees (Table 5). Decreasing
the cut-point to 0.20 could increase the correct prediction of
diseased trees to 84.0% but the ethanol concentration associated
with this cut-point was less distinct than the one for 0.40. Table 6
provides further details regarding predictions with cut-point 0.40
relative to where the stain-infection was observed. There were
10 trees identified as infected without visual confirmation (false
positives), but we suspect some portion of them were infected,
especially those with the highest ethanol concentrations.
Ethanol concentrations indicate that Douglas-fir roots infected
with P. sulphurascens experience the most stress in cells nearest
the infection zone. Lower ethanol levels further away results from
the multiple dissipation processes, or sinks. First, ethanol hydrogen
bonds to water (Fileti et al., 2004) and slowly diffuses in all directions into surrounding tissues. Second, it can enter xylem tracheids
and be transported more rapidly upward and over greater distances in the transpiration stream as shown by flooding roots
(Kreuzwieser et al., 1999; Cojocariu et al., 2004). This likely contributed to the correlation (r = 0.729) between concentrations in
roots and lower bole in diseased trees at site 1, and to some of
268
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
Pre-dawn water potential (MPa)
0.0
AA
Table 3
Sites 1, 2 statistical analysis results for pre-dawn water potentials in diseased and
healthy trees.
-0.5
a
a
a
-1.0
a
-1.5
b
-2.0
Effecta
d.f.
F
Pb
Tc
M
Tc M
1, 45.9
1, 45.9
1, 45.9
1.18
276.99
2.33
0.282
<0.001
0.134
Tc = tree condition, M = month.
Bold P values are statistically significant.
Table 4
Sites 3, 4 statistical analysis results for ethanol, acetone, and water contents in root
phloem and xylem from diseased and healthy trees.
-2.5
Healthy
Diseased
Volatile
Effecta
Tree condition
Pre-dawn water potential (MPa)
0.0
Ethanol
Acetone
Water
B
b
-0.5
a
b
Tc
Tc
Tc
Phloem
Xylem
d.f.
F
Pb
d.f.
F
Pb
1, 78.0
1, 78.0
1, 77.8
2.54
13.61
0.25
0.115
<0.001
0.617
1, 78.0
1, 73.1
1, 78.0
23.84
41.17
18.95
<0.001
<0.001
<0.001
Tc = tree condition.
Bold P values are statistically significant.
-1.0
healthy trees at sites 1, 2 for one or more sampling dates, none
were selected as model parameters and therefore not presented
for trees at sites 3, 4, or the root transects.
a
-1.5
4.3. Water and crown symptoms
-2.0
-2.5
Sept
Nov
Month
Fig. 4. Sites 1, 2 predawn (95% CI) water potentials in Douglas-fir for the main
effects of tree condition (A), month sampled (B). Bars with the same letters are not
statistically different, see Table 3 for P values.
the seasonal decline in xylem ethanol of diseased tree roots at sites
1, 2, because Douglas-fir sap flow and transpiration are sensitive to
soil moisture and vapor pressure deficit (Link et al., 2014). High
September ethanol concentrations coincided with dry soils that
received only 3.0 cm of rainfall the previous three months. By
November, sap flow would have increased in response to 43.6 cm
of rainfall the previous month, causing ethanol concentrations to
decrease. The subsequent 125.3 cm of winter and spring precipitation allowed extended periods of high sap flow further reducing
ethanol to the low May quantities. Metabolism into other cellular
constituents (MacDonald and Kimmerer, 1993) or acetaldehyde
(Kreuzwieser et al., 1999; Cojocariu et al., 2004; Tohmuram et al.,
2012) upon entering healthy live cells is the third major sink.
Finally, low ethanol concentrations in the infected tissue may
result in part from metabolism by the pathogen, as demonstrated
for another tree pathogen, Armillaria mellea (Weinhold and
Garraway, 1966).
4.2. Acetone and other headspace volatile concentrations
Acetone concentrations were the opposite of ethanol, with
highest quantities within the infected-stained area. Thus, samples
must contain infected tissue for acetone to be a useful disease indicator. Furthermore, the magnitudes of differences between diseased and healthy tissue were much smaller than ethanol and
therefore a less sensitive biomarker. Acetone is generated from tissue components during sample heating (Kelsey et al., 1998b) and is
not subject to concentration changes from diffusion or transport in
water. Although acetaldehyde, methanol, and propanol concentrations were higher in one or more tissues of diseased trees than
Water measurements are not useful for detecting trees infected
with this pathogen. Height growth of Douglas-fir and other conifers is more sensitive and impaired earlier than radial growth
when stressed for water (Rais et al., 2014; Klein et al., 2011), so
reduced height growth would be expected if P. sulphurascens infections decrease crown water supplies sufficiently. But this appears
only in the most advanced disease stages (Bloomberg and Wallis,
1979; Thies, 1983), with no external signs of loss in vigor until after
the fungus grows through the bole sapwood (Buckland et al.,
1954). Adequate crown water supply is also indicated by trees
without crown symptoms blown down during storms although
they have extensive root decay. Thies (1983) report no correlation
between impaired tree growth and the number, size, or percentage
of infected roots (Thies, 1983).
4.4. Model prediction of trees with laminated root rot
The model using xylem ethanol concentrations performed best
for predicting trees with infection observed in roots and the root
collar because they have a larger infected area with high ethanol.
The models limited ability to detect trees with stain only at the
root collar (stump surface) most likely resulted because their
infected roots were deeper in the soil or beneath the root collar
and not sampled. In some trees the lower bole heartwood is
decayed, or rotted away, but otherwise surrounded by healthy sapwood with little or no infection in lateral roots (Thies and
Westlind, 2005). Root cores would not detect this type of infection.
Overall the model performed well given the three dimensional root
structures (McMinn, 1963; Mauer and Palátová, 2012) and the limited size and number of samples per tree.
High ethanol outside the stained-infection area allows diseased
root detection without sampling infected tissue, so some of the
false positive trees were likely diseased. Although the model was
created with measurements from 1 cm depth xylem cores, it identified diseased trees at sites 3, 4 using 6 cm depth cores suggesting
that core depth in the trees sampled here was not influencing tissue ethanol concentration enough to strongly reduce its value for
predicting disease. Core depth, root size, and infection position
269
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
14
12
10
a
8
6
a
4
2
0
Xylem B
Ethanol, µg.g-1 fresh mass
100
50
10
b
5
a
c
0
Healthy
Diseased
Tree condition
Fig. 5. Sites 3, 4 backtransformed mean ethanol (95% CI) concentrations in Douglasfir root phloem (A) and xylem (B) of diseased and healthy trees. Bars with the same
letters are not statistically different. See Table 4 for P values. Note the different Y
scales and break in xylem graph B.
all interact to influence the measured ethanol concentrations. For
example, deep cores from large roots increase the likelihood of
reaching tissues with higher ethanol concentrations, but may
increase healthy tissue proportions with lower ethanol, causing a
dilution effect.
4.5. Ethanol use for hazard tree identification and with extensive
ground surveys
Ethanol analysis as done in this study is a useful research tool,
but would have limited use for improving hazard tree assessments
since elevated concentrations typically occur close to infections.
Root coring and visual inspection for stain or decay is adequate
for identifying hazard trees with this disease. Decaying roots with
soft tissue are easily recognized when the increment borer spins or
stops moving forward. Roots here were sampled 5–30 cm beyond
their entry point into the litter layer, but for hazard trees a set distance may be better. Sampling closer to the root collar can decrease
root excavation time, but would miss some roots with less
advanced infections. Battery powered drills quickly extract cores
and need to penetrate the entire root diameter to avoid missing
any infection. The number of roots sampled per tree can be
adjusted to match the potential damages should the tree fall, but
four major roots; one per cardinal direction seems a minimum.
Sides with no accessible roots can be cored near the root collar into
the bole center as a substitute. Drilling slightly above the root collar minimizes twisted cores sticking in the borer, and will help
detect trees with decay restricted in their centers. The number,
size, and location of roots with decay determine whether the tree
is sufficiently hazardous to warrant immediate removal.
Incorporating ethanol analysis as conducted here into extensive
ground surveys has two major limitations; (1) the need for
multiple samples per tree and (2) laboratory analysis with expensive instruments, or a considerable time requirement. As discussed
later, we propose that ethanol release to the atmosphere contributes to bark beetle host selection in trees with advanced disease. Whether trees with less infection release sufficient ethanol
to confirm the disease presence remains to be determined. If so,
noninvasive field sensors could have value detecting them if they
provide rapid, accurate detection and are easily portable under
the harsh and variable conditions of natural environments. There
are various technologies with potential to meet these criteria with
further research and development, including proton transfer reaction mass spectrometry (PTR-MS; Ellis and Mayhew, 2014;
Holzinger et al., 2000; Rottenberger et al., 2008; Kaser et al.,
2013), intelligent electronic nose systems (Baietto et al., 2010;
Naher et al., 2013), wireless smart phone sensors (Azzarelli et al.,
2014), and handheld analyzers used to measure ethanol in human
breath (Workman, 2012; Andrews, 2013). Dogs are also a possibility as they can detect rot in chemically treated utility poles
(Davner, 1986).
4.6. Ethanol’s potential role in primary beetle attraction and host
selection
Standing Douglas-fir infected with P. sulphurascens are more
vulnerable to Douglas-fir beetle, Dendroctonus pseudosugae Hopkins, attack than adjacent healthy trees (Goheen and Hansen,
1993), but not until the fungus has grown through the lower bole
sapwood (Buckland et al., 1954). Diseased true firs, Abies spp., are
typically colonized by the fir engraver, Scolytus ventralis LeConte,
and sometime Dryocetes confusus Swaine, but only after extensive
root crown infection (Lane and Goheen, 1979). This late bark beetle
Phloem A
Acetone, µg.g-1 fresh mass
Phloem A
8
b
6
4
a
2
0
Xylem B
Acetone, µg.g-1 fresh mass
Ethanol, µg.g-1 fresh mass
16
8
6
4
b
2
a
0
Healthy
c
Diseased
Tree condition
Fig. 6. Sites 3, 4 backtransformed mean acetone (95% CI) concentrations in
Douglas-fir phloem (A) and xylem (B) of diseased and healthy trees. Bars with the
same letters are not statistically different. See Table 4 for P values.
270
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
Table 5
Classification table based on the ethanol logistic regression model using a cut-point
probability of 0.40. The overall rate of correct classification for both uninfected and
infected trees = ([45 + 17]/80) 100 = 77.5%; overall rate of correctly predicting the
infected trees only = (17/25) 100 = 68.0%.
Observed
Model predictions
Uninfected P(i) < 0.4a
Infected P(i) > 0.4
a
Uninfected
55
Infected
25
Total
80
45
10
8
17
53
27
Probability of infection.
arrival may result from their inability to detect trees with less severe disease, a slow weakening of the tree’s chemical defenses, or
some combination. Using the new ethanol information here with
previous literature reports for ethanol, pathogen biology, and
Douglas-fir beetle primary attractants allows us to propose a
mechanism of how stress induced ethanol may play a central role
in bark beetle primary host selection of trees infected with this
pathogen.
Release of bole generated primary attractants is consistent with
the absence of beetle attacks on wind thrown trees with severe
root infection. The pathogens infection behavior is an important
factor because the mycelium grows upward through root xylem
but initially enters only the lower bole heartwood near the sapwood border. It spreads along this boundary fusing with adjacent
mycelium to form an arc shaped band, or circle of infection
(Buckland et al., 1954; Thies and Sturrock, 1995). From this interior
position the pathogen slowly grows vertically and radially through
the bole sapwood, but with minimal impact on sap flow as shown
with water potential measurements. When the lower bole sapwood becomes thoroughly infected vigor starts to decline, crown
symptoms appear (Bloomberg and Wallis, 1979) and eventually
bark beetles attack (Buckland et al., 1954).
As the infection progresses from roots to lower bole there are
two sources contributing to elevated ethanol in the bole tissues.
The first is ethanol synthesized in stressed roots and transported
by sap flow before the pathogen enters the root collar. This would
contribute to the strong correlation observed between lower bole
and root concentrations in sapwood (r = 0.729) of diseased trees
at site 1. But the absence of beetle attacks before bole infection
suggests this ethanol supply is not adequate for their detection,
probably because of the multiple internal sinks limiting accumulation. A portion of the ethanol entering the lower bole by sap flow is
transient and will continue moving upward while another portion
will diffuse into surrounding tissue diluting the concentration,
including some being hydrogen bound in cellulose and other cell
wall polymer (Stamm and Tarkow, 1950; Mantanis et al., 1995)
temporarily holding it in place. Any diffusing radially outward will
enter healthy cambium and phloem where it can be metabolized
(MacDonald and Kimmerer, 1993). These sinks limit the quantity
escaping through the outer bark. Low ethanol levels in Douglasfir boles has been proposed as an explanation for the absence of
Douglas-fir beetle attacks on trees severely stressed by foliar infection of Swiss needle cast (Kelsey and Manter, 2004). Roots with
advanced infection may also release some ethanol into surrounding soils but what portion can escape above ground is unknown.
It is likely minimal from roots beneath the bole or deeply buried
because of soil water absorption and uptake by healthy tree roots
(Joseph and Kelsey, 2000) or other plant roots, hydrogen bonding
to any cellulosic matter or charcoal (Stamm and Tarkow, 1950;
Mantanis et al., 1995; Kelsey et al., 2013), or a portion might be
metabolized by microbes, as in live tree cells (MacDonald and
Kimmerer, 1993).
A second source of bole ethanol can be initiated after the pathogen begins stressing live xylem cells in the inner most sapwood.
However, there is no immediate change in ethanol release at the
bark surface for two reasons. First, the inner bole xylem has a limited live cell volume (Stockfors and Linder, 1998) and produces
much lower quantities of ethanol than the roots and phloem
(Kelsey et al., 1998a) or cambium (Kimmerer and Stringer, 1988).
Second, its accumulation there will be strongly influenced by the
three sinks mentioned above. The abundance of surrounding tissue
allows dilution by diffusion in any direction, movement radially
across the sapwood is vulnerable to unhindered sap flow transport
upward, and any entering healthy cambium and phloem is at risk
of metabolism. However, these factors all change radically once
the pathogen spreads through the sapwood and begins stressing
cambium and phloem. First, ethanol synthesis increases because
both tissues have greater live cell volume (Stockfors and Linder,
1998) and can produce more ethanol than sapwood (Kelsey et al.,
1998a; Kimmerer and Stringer, 1988). Second, the radial distance
from site of synthesis to the atmosphere is shorter and the accompanying sink strengths decline. While diffusion can still occur in all
directions, it no longer passes through sapwood, eliminating the
sap flow sink. Also, as the pathogen penetrates cambium and
phloem they most likely lose some ability to metabolize ethanol
and this sink also declines. At some point ethanol release to the
atmosphere increases as demonstrated by trapping it on the bark
surface after infusing sapwood in bole segments (Kelsey et al.,
2013), from boles with flooded roots (Ranger et al., 2013), or lodgepole pine stems with active heartwood decay fungi (Gara et al.,
1993).
Bole released ethanol mixed with Douglas-fir monoterpenes, or
resin will be a strong primary attractant for D. pseudosugae (Jantz
and Rudinsky, 1966; Rudinsky, 1966; Pitman et al., 1975). Fresh
Douglas-fir monoterpenes or oleoresins released alone function
as primary attractants for this beetle (Rudinsky, 1966; Johnson
Table 6
Ethanol logistic regression model predictions at cut-point probability of 0.40 sorted by observed stain locations.
Model predictions:
Observed stain location
No. trees (observed)
%
Infected:
Correct
Incorrect, false negative
Correct
Incorrect, false negative
Correct
Incorrect, false negative
Stump surface
Stump surface
Core sample
Core sample
Stump and core
Stump and core
(25)
1/7
6/7
5/7
2/7
11/11
0/11
14.3
85.7
71.4
28.5
100.0
0
Not infected:
Correct
Incorrect, false positive?a
No stain
No stain
(55)
45/55
10/55
81.8
18.2
a
These may or may not be false positive trees as one of their root cores may have been near infected tissue that was not observed. Additional root sampling or excavation
would be needed to verify infection.
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
and Belluschi, 1969; Pureswaran and Borden, 2005), but sufficient
quantities may not be released alone from P. sulphurascens infected
Douglas-fir as there is no external resin leakage or resinosis prior to
beetle attacks. Ethanol alone can also function as a primary attractant for D. pseudosugae (Stoszek, 1973; Pitman et al., 1975), but
host monoterpenes are always present and their release rates
may increase with ethanol release rates, as demonstrated for
healthy and diseased lodgepole pine (Gara et al., 1993). Pioneering
D. pseudosugae attacks need not start at the root collar given their
preference for positions higher on the bole (Furniss and Kegley,
2014) and their unfocused response to attractants that results in
attacks on unbaited trees 15 m or more from baited trees
(Rudinsky, 1966; Johnson and Belluschi, 1969; Thier and
Patterson, 1997). Any successful attacks by pioneering beetles will
quickly initiate stronger secondary attraction when monoterpene
and ethanol vapors mix with pheromones (Pitman et al., 1975;
Ross and Daterman, 1995). We suspect ethanol mixed with
monoterpenes also attract S. ventralis to trees infected with P. sulphurascens given the absence of known pheromone-mediated secondary attraction for this bark beetle (Macías-Sáman et al., 1998).
Additional experiments on ethanol and monoterpene emissions
from P. sulphurascens infected trees are needed to further validate
this proposed mechanism.
4.7. Potential changes in chemical defense against bark beetles
When the first pioneering D. pseudosugae beetle(s) initiate their
attack on P. sulphurascens stressed Douglas-fir the quality and
quantity of constitutive oleoresin, followed by induced oleoresin
flow, function as the critical chemical defense (Jantz and
Rudinsky, 1966; Rudinsky, 1966), as for other resinous conifers
(Franceschi et al., 2005). In general, a reduction in tree vigor is
accompanied by reduced resin flow (Fettig et al., 2007). In southwestern ponderosa pine, resin flow declines in conjunction with
reduced basal area increment growth (BAI) (McDowell et al.,
2007), because the number of resin ducts correlate positively with
BAI growth (Kane and Kolb, 2010). The latter investigators found
trees surviving drought-associated bark beetle attacks had larger
resin ducts, at higher densities and a greater proportion of the
annual growth than in dead trees. Douglas-fir resin flow also declines when BAI decreases in response to needle loss and infection
from the Swiss needle cast fungus, Phaeocryptopus gaeumannii
(Rhode) (Kelsey and Manter, 2004). However, since Douglas-fir
infected with P. sulphurascens do not show changes in radial
growth or water status, both factors that can directly influence
resin flow for short periods (Lorio, 1994; Fettig et al., 2007), until
after bole sapwood is thoroughly infected, we suspect minimal
changes in constitutive or induced resin flow chemical defense
until the most advanced stages of infection. Pioneering Douglasfir beetles will attack healthy trees with unimpaired resin flow
defenses when host primary attractants are fresh enough
(Johnson and Belluschi, 1969). We suspect enhanced release of
ethanol and monoterpenes may stimulate them to attack P. sulphurascens prior to substantial reductions in constitutive or
induced resin flow defense, but such attacks may not be successful
until there is sufficient loss of tree vigor to further impair these
defenses. Even if these defenses do mitigate successful tree colonization, it is highly likely that ethanol release in conjunction with
host monoterpenes drive the initial host finding and selection of P.
sulphurascens stressed trees by pioneering beetles.
Acknowledgements
We thank Starker Forests, Inc., Craig Olsen, Greg Johnson, and
Christie Shaw for assistance in establishing this research by
helping to locate suitable sites, and for access to trees on private
271
property they managed. Funding was provided by USDA Forest Service, Pacific Northwest Research Station. The use of trade, firm, or
corporation names is for information and convenience of the
reader and does not constitute an official endorsement or approval
by the U.S. Department of Agriculture. We also thank Greg Filip,
Iral Ragenovich, José Negrón and Ariel Muldoon for their reviews
and helpful suggestions.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.foreco.2015.10.
039.
References
Andrews, T.M., 2013. Breathalyzers of the Future Today. The Atlantic. <http://www.
theatlantic.com/health/archive/2013/06/breathalyzers-of-the-future-today/
277249/> (accessed 01.06.15).
Azzarelli, J.M., Mirica, K.A., Ravnsbaek, J.B., Swager, T.M., 2014. Wireless gas
detection with a smartphone via rf communication. Proc. Natl. Acad. Sci. 111,
18162–18166. http://dx.doi.org/10.1073/pnas.1415403111.
Baietto, M., Wilson, A.D., Bassi, D., Ferrini, F., 2010. Evaluation of three electronic
noses for detecting incipient wood decay. Sensors 10, 1062–1092. http://dx.doi.
org/10.3390/s100201062.
Bloomberg, W.J., Reynolds, G., 1982. Factors affecting transfer and spread of
Phellinus weirii mycelium in roots of second-growth Douglas-fir. Can. J. For. Res.
12, 424–427. http://dx.doi.org/10.1139/x82-062.
Bloomberg, W.J., Wallis, G.W., 1979. Comparison of indicator variables for
estimating growth reduction associated with Phellinus weirii root rot in
Douglas-fir plantations. Can. J. For. Res. 9, 76–81. http://dx.doi.org/10.1139/
x79-013.
Buckland, D.C., Molnar, A.C., Wallis, G.W., 1954. Yellow laminated root rot of
Douglas fir. Can. J. Bot. 32, 69–81. http://dx.doi.org/10.1139/b54-009.
Cojocariu, C., Kreuzwieser, J., Rennenberg, H., 2004. Correlation of short chained
carbonyls emitted from Picea abies with physiological and environmental
parameters. New Phytol. 162, 717–727. http://dx.doi.org/10.1111/j.14698137.2004.01061.x.
Cook, R.J., Edmonds, R.L., Klopfenstein, N.B., Littke, W., McDonald, G., Omdal, D.,
Ripley, K., Shaw, C.G., Sturrock, R., Zambino, P., 2013. Opportunities for
Addressing Laminated Root Rot Caused by Phellinus sulphurascens in
Washington’s Forests. A Report from the Washington State Academy of
Sciences in Cooperation with the Washington State Department of Natural
Resources, Olympia, WA, 110 p. <http://www.washacad.org/>.
Davner, L., 1986. Vilken röta! Hundar I jakten på skadade träd. Skogen 5, 48.
Ellis, A.M., Mayhew, C.A., 2014. Proton Transfer Reaction Mass Spectrometry:
Principles and Applications. John Wiley & Sons Ltd., Chichester, UK. http://dx.
doi.org/10.1002/9781118682883.fmatter.
Fettig, C.J., Klepzig, K.D., Billings, R.F., Munson, A.S., Nebeker, T.E., Negrón, J.F.,
Nowak, J.T., 2007. The effectiveness of vegetation management practices for
prevention and control of bark beetle infestations in coniferous forests of the
western and southern United States. For. Ecol. Manage. 238, 24–53. http://dx.
doi.org/10.1016/j.foreco.2006.10.011.
Fileti, E.E., Chaudhuri, P., Canuto, S., 2004. Relative strength of hydrogen bond
interaction in alcohol–water complexes. Chem. Phys. Lett. 400, 494–499. http://
dx.doi.org/10.1016/j.cplett.2004.10.149.
Franceschi, V.R., Krokene, P., Christiansen, E., Kreking, T., 2005. Anatomical and
chemical defenses of conifer bark against bark beetles and other pests. New
Phytol. 167, 353–376. http://dx.doi.org/10.1111/j.1469-8137.2005.01436.x.
Furniss, M.M., Kegley, S.J., 2014. Douglas-fir Beetle. USDA For. Serv. For. Insect Dis.
Leaflet 5.
Gara, R.I., Littke, W.R., Rhoades, D.F., 1993. Emission of ethanol and monoterpenes
by fungal infected lodgepole pine trees. Phytochemistry 34, 987–990. http://dx.
doi.org/10.1016/S0031-9422(00)90699-X.
Goheen, D.J., Hansen, E.M., 1993. Effects of pathogens and bark beetles on forests.
In: Schowalter, T.B., Filip, G.M. (Eds.), Beetle-Pathogen Interactions in Conifer
Forests. Academic Press, London and New York, NY, pp. 175–196.
Holzinger, R., Sandoval-Soto, L., Rottenberger, S., Crutzen, P.J., Kesselmeier, J., 2000.
Emissions of volatile organic compounds from Quercus ilex L. measured by
proton transfer reaction mass spectrometry under different environmental
conditions. J. Geophys. Res. 105 (D16), 20573–20579. http://dx.doi.org/10.1029/
2000JD900296.
Hosmer Jr., D.W., Lemeshow, S., Sturdivant, R.X., 2013. Applied Logistic Regression,
vol. 398. John Wiley & Sons, Hoboken, New Jersey.
Jantz, O.K., Rudinsky, J.A., 1966. Studies of the Olfactory Behavior of the Douglas-fir
Beetle, Dendroctonus pseudotsugae Hopkins. Oregon State Univ. Tech. Bull. 94.
Johnson, N.E., Belluschi, P.G., 1969. Host-finding behavior of the Douglas-fir beetle.
J. For. 67, 290–295.
Joseph, G., Kelsey, R.G., 2000. Physiology and growth of Douglas-fir seedlings
treated with ethanol solutions. Plant Sci. 150, 191–199. http://dx.doi.org/
10.1016/S0168-9452(99)00189-2.
272
R.G. Kelsey et al. / Forest Ecology and Management 360 (2016) 261–272
Joseph, G., Kelsey, R.G., 2004. Ethanol synthesis and aerobic respiration in the
laboratory by leader segments of Douglas-fir seedlings from winter and spring.
J. Exp. Bot. 55, 1095–1103. http://dx.doi.org/10.1093/jxb/erh116.
Kane, J., Kolb, T.E., 2010. Importance of resin ducts in reducing ponderosa pine
mortality from bark beetle attacks. Oecologia 164, 601–609. http://dx.doi.org/
10.1007/s00442-010-1683-4.
Kaser, L., Karl, T., Schnitzhofer, R., Graus, M., Herdlinger-Blatt, I.S., DiGangi, J.P., Sive,
B., Turnipseed, A., Hornbrook, R.S., Zheng, W., Flocke, F.M., Guenther, A.,
Keutsch, F.N., Apel, E., Hansel, A., 2013. Comparison of different real time VOC
measurement techniques in a ponderosa pine forest. Atmos. Chem. Phys. 13,
2893–2906. http://dx.doi.org/10.5194/acp-13-2893-2013.
Kelsey, R.G., Joseph, G., 1998. Ethanol in Douglas-fir with black-stain root disease
(Leptographium wageneri). Can. J. For. Res. 28, 1207–1212. http://dx.doi.org/
10.1139/x98-087.
Kelsey, R.G., Joseph, G., 2001. Attraction of Scolytus unispinosus bark beetles to
ethanol in water-stressed Douglas-fir branches. For. Ecol. Manage. 144, 229–
238. http://dx.doi.org/10.1016/S0378-1127(00)00387-X.
Kelsey, R.G., Joseph, G., 2003. Ethanol in ponderosa pine as an indicator of
physiological injury from fire and its relationship to secondary beetles. Can. J.
For. Res. 33, 870–884. http://dx.doi.org/10.1139/X03-007.
Kelsey, R.G., Manter, D.K., 2004. Effect of Swiss needle cast on Douglas-fir stem
ethanol and monoterpene concentrations, oleoresin flow, and host selection by
the Douglas-fir beetle. For. Ecol. Manage. 190, 241–253. http://dx.doi.org/
10.1016/j.foreco.2003.10.019.
Kelsey, R.G., Joseph, G., Gerson, E.A., 1998a. Ethanol synthesis, nitrogen,
carbohydrates, and growth in tissues from nitrogen fertilized Pseudotsuga
menziesii (Mirb.) Franco and Pinus ponderosa Dougl. ex Laws. seedlings. Trees 13,
103–111. http://dx.doi.org/10.1007/PL00009738.
Kelsey, R.G., Joseph, G., McWilliams, M.G., 2011. Ethanol synthesis by anoxic root
segments from five cedar species relates to their habitat attributes but not their
known differences in vulnerability to Phytophthora lateralis root disease. Can. J.
For. Res. 41, 1202–1211. http://dx.doi.org/10.1139/X11-043.
Kelsey, R.G., Joseph, G., Thies, W.G., 1998b. Sapwood and crown symptoms in
ponderosa pine infected with black-stain and annosum root disease. For. Ecol.
Manage. 111, 181–191. http://dx.doi.org/10.1016/S0378-1127(98)00332-6.
Kelsey, R.G., Thies, W.G., Schmitt, C.L., 2006. Using chemicals to detect root disease
in ponderosa pine stands with a low incidence of disease in eastern Oregon. For.
Ecol. Manage. 232, 205–215. http://dx.doi.org/10.1016/j.foreco.2006.05.073.
Kelsey, R.G., Beh, M.M., Shaw, D.C., Manter, D.K., 2013. Ethanol attracts scolytid
beetles to Phytophthora ramorum cankers on coast live oak. J. Chem. Ecol. 39,
494–506. http://dx.doi.org/10.1139/X11-043.
Kelsey, R.G., Gallego, D., Sánchez-García, F.J., Pajares, J.A., 2014. Ethanol
accumulation during severe drought may signal tree vulnerability to
detection and attack by bark beetles. Can. J. For. Res. 44, 554–561. http://dx.
doi.org/10.1139/cjfr-2013-0428.
Kimmerer, T.W., Stringer, M.A., 1988. Alcohol dehydrogenase and ethanol in the
stems of trees. Plant Physiol. 87, 693–697. http://dx.doi.org/10.1104/
pp.87.3.693.
Klein, T., Cohen, S., Yakir, D., 2011. Hydraulic adjustments underlying drought
resistance of Pinus halepensis. Tree Physiol. 31, 637–648. http://dx.doi.org/
10.1093/treephys/tpr047.
Kolb, B., Pospisil, P., Auer, M., 1984. Quantitative headspace analysis of solid
samples: a classification of various sample types. Chromatographia 19, 113–
122. http://dx.doi.org/10.1007/BF02687726.
Kreuzwieser, J., Scheerer, U., Rennenberg, H., 1999. Metabolic origin of acetaldehyde
emitted by poplar (Populus tremula P. alba) trees. J. Exp. Bot. 50, 757–765.
http://dx.doi.org/10.1093/jxb/50.335.757.
Lane, B.B., Goheen, D.J., 1979. Incidence of root disease in bark beetle-infested
eastern Oregon and Washington true firs. Plant Dis. Rep. 63, 262–266.
Link, P., Simonin, K., Maness, H., Oshun, J., Dawson, T., Fung, I., 2014. Species
differences in the seasonality of evergreen tree transpiration in a Mediterranean
climate: analysis of multiyear, half-hourly sap flow observations. Water Resour.
Res. 50, 1869–1894. http://dx.doi.org/10.1002/2013WR014023.
Littell, R.C., Milliken, G.A., Stroup, W.W., Wolfinger, R.D., Schabenberger, O., 2006.
SASÒ for Mixed Models, second ed. SAS Institute Inc., Cary, NC.
Lorio Jr., P.L., 1994. The relationship of oleoresin exudation pressure (or lack
thereof) to flow from wounds. J. Sustain. For. 1, 81–93. http://dx.doi.org/
10.1300/J091v01n04_04.
MacDonald, R.C., Kimmerer, T.W., 1993. Metabolism of transpired ethanol by
eastern cottonwood (Populus deltoides Bartr.). Plant Physiol. 102, 173–179.
http://dx.doi.org/10.1104/pp.102.1.173.
Macías-Sáman, J.E., Borden, J.H., Gries, R., Pierce Jr., H.D., Gries, G., King, G.G.S., 1998.
Primary attraction of the fir engraver, Scolytus ventralis. J. Chem. Ecol. 24, 1049–
1075. http://dx.doi.org/10.1023/A:1022354503443.
Mantanis, G.I., Young, R.A., Rowell, R.M., 1995. Swelling of compressed cellulose
fiber webs in organic liquids. Cellulose 2, 1–22. http://dx.doi.org/10.1007/
BF00812768.
Mauer, O., Palátová, E., 2012. Root system development in Douglas fir (Pseudotsuga
menziesii [Mirb.] Franco) on fertile sites. J. For. Sci. 58, 400–409.
McCauley, K.J., Cook, S.A., 1980. Phellinus weirii infestation of two mountain
hemlock forests in the Oregon Cascades. For. Sci. 26, 23–29.
McDowell, N.G., Adams, H.G., Bailey, J.D., Kolb, T.E., 2007. The role of stand density
on growth efficiency, leaf area index, and resin flow in southwestern ponderosa
pine forests. Can. J. For. Res. 37, 343–355. http://dx.doi.org/10.1139/X06-233.
McMinn, R.G., 1963. Characteristics of Douglas-fir root systems. Can. J. Bot. 41, 105–
122. http://dx.doi.org/10.1139/b63-010.
Naher, L., Yusuf, U.K., Ismail, A., Tan, S.G., Mondal, M.M.A., 2013. Ecological status of
Ganoderma and basal stem rot disease of oil palms (Elaeis guineensis Jacq.). Aust.
J. Crop Sci. 7, 1723–1727.
Nelson, E.E., Hartman, T., 1975. Estimating spread of Poria weirii in a high elevation
mixed conifer stand. J. For. 73, 141–142.
Pitman, G.B., Hedden, R.L., Gara, R.I., 1975. Synergistic effects of ethyl alcohol on the
aggregation of Dendroctonus pseudotsugae (Col., Scolytidae) in response to
pheromones. Z. Angew. Entomol. 78, 203–208.
Pureswaran, D.P., Borden, J.H., 2005. Primary attraction and kairomonal host
discrimination in three species of Dendroctonus (Coleoptera: Scolytidae). Agric.
For. Entomol. 7, 219–230. http://dx.doi.org/10.1111/j.1461-9555.2005.00264.x.
Ranger, C.M., Reding, M.E., Schultz, P.B., Oliver, J.B., 2013. Influence of flood-stress
on ambrosia beetle host-selection and implications for their management in a
changing climate. Agric. For. Entomol. 15, 56–64. http://dx.doi.org/10.1111/
j.1461-9563.2012.00591.x.
Rais, A., van de Kuilen, J.-W.G., Pretzsch, H., 2014. Growth reaction patterns of tree
height, diameter, and volume of Douglas-fir (Pseudotsuga menziesii [Mirb.]
Franco) under acute drought stress in southern Germany. Eur. J. For. Res. 133,
1043–1056. http://dx.doi.org/10.1007/s10342-014-0821-7.
Ross, D.W., Daterman, G.E., 1995. Response of Dendroctonus pseudotsugae
(Coleoptera: Scolytidae) and Thanasimus undatulus (Coleoptera: Cleridae) to
traps with semiochemicals. J. Econ. Entomol. 88, 106–111. http://dx.doi.org/
10.1093/jee/88.1.106.
Rottenberger, S., Kleiss, B., Kuhn, U., Wolf, A., Piedade, M.T.F., Junk, W., Kesselmeier,
J., 2008. The effect of flooding on the exchange of the volatile C2-compounds
ethanol, acetaldehyde and acetic acid between leaves of Amazonian floodplain
tree species and the atmosphere. Biogeosciences 5, 1085–1100. http://dx.doi.
org/10.5194/bg-5-1085-2008.
Rudinsky, J.A., 1966. Host selection and invasion by the Douglas-fir beetle,
Dendroctonus pseudostugae Hopkins, in coastal Douglas-fir forests. Can.
Entomol. 98, 98–111. http://dx.doi.org/10.4039/Ent9898-1.
SAS Institute Inc., 2012. SAS/STATÒ 9.3. User’s Guide. SAS Institute Inc., Cary, North
Carolina, USA.
Stamm, A.J., Tarkow, H., 1950. Penetration of cellulose fibers. J. Phys. Coll. Chem. 54,
745–753. http://dx.doi.org/10.1021/j150480a001.
Stoszek, K.J., 1973. A Contribution to the Biology of the Pseudohylesinus nebulosus
(LeConte) (Coleoptera:Scolytidae), Especially in Relation to the Moisture Stress
of its Host, Douglas-fir. Ph.D. Dissertation. Oregon State University, Corvallis.
Stockfors, J., Linder, S., 1998. Effect of nitrogen on the seasonal course of growth and
maintenance respiration in stems of Norway spruce trees. Tree Physiol. 18,
155–166. http://dx.doi.org/10.1093/treephys/18.3.155.
Thier, R.W., Patterson, S., 1997. Mortality of Douglas-fir after operational
semiochemical baiting for Douglas-fir beetle (Coleoptera: Scolytidae). West. J.
Appl. For. 12, 16–20.
Thies, W.G., 1983. Determination of growth reduction in Douglas-fir infected by
Phellinus weirii. For. Sci. 29, 305–315.
Thies, W.G., Nelson, E.E., 1997. Laminated root rot: new considerations for surveys.
West. J. Appl. For. 12, 49–51.
Thies, W.G., Sturrock, R.N., 1995. Laminated Root Rot in Western North America.
USDA For. Serv. Res. Sta. PNW-GTR-349.
Thies, W.G., Westlind, D.J., 2005. Stump removal and fertilization of five Phellinus
weirii-infested stands in Washington and Oregon affect mortality and growth of
planted Douglas-fir 25 years after treatment. For. Ecol. Manage. 219, 242–258.
http://dx.doi.org/10.1016/j.foreco.2005.08.050.
Tohmuram, S.-I., Ishikawa, A., Miyamoto, K., Inoue, A., 2012. Acetaldehyde emission
from wood induced by the addition of ethanol. J. Wood Sci. 58, 57–63. http://dx.
doi.org/10.1007/s10086-011-1215-9.
Wallis, G.W., Bloomberg, W.J., 1981. Estimating the total extent of Phellinus weirii
root rot centers using above- and below-ground disease indicators. Can. J. For.
Res. 11, 827–830. http://dx.doi.org/10.1139/x81-118.
Wallis, G.W., Reynolds, G., 1965. The initiation and spread of Poria weirii root rot of
Douglas fir. Can. J. Bot. 43, 1–9. http://dx.doi.org/10.1139/b65-001.
Weinhold, A.R., Garraway, M.O., 1966. Nitrogen and carbon nutrition Armillaria
mellea in relation to growth-promoting effects of ethanol. Phytopathol. 56, 108–
112.
Workman Jr., T.E., 2012. The science behind breath testing for ethanol. Univ. Mass.
Law Rev. 7, 110–180.